Dreadnode is a privately held cybersecurity company founded in 2023 that specializes in offensive AI security. The company’s mission is to advance the state of offensive security by developing AI systems and tools that match or exceed human capabilities in identifying and exploiting vulnerabilities in digital systems. Dreadnode’s platform enables enterprises, governments, and AI developers to safely simulate, test, and evaluate AI models and applications against realistic attack scenarios. By focusing on offensive machine learning, Dreadnode aims to equip organizations with the means to proactively identify and mitigate security risks in AI deployments, fostering a future where autonomous AI can be trusted to perform offensive tasks securely and ethically.
Leadership
Will Pearce (Co-founder & CEO): Former lead of NVIDIA’s AI red team, with extensive hands-on experience in offensive security, penetration testing, and AI-driven adversarial research. Pearce has also contributed to security initiatives at Microsoft and other major technology firms.
Nick Landers (Co-founder & CTO): Previously VP of Research at NetSPI, Landers has a background in building and augmenting AI red teams at Microsoft, NVIDIA, and Fortune 500 companies. He is recognized for his expertise in adversarial machine learning and red teaming offensive security tools.
Core Technologies
Dreadnode’s core technologies are designed to simulate, test, and evaluate AI systems from an offensive security perspective:
Strikes: An AI agent training ground focused on offensive cybersecurity, providing real-world scenarios to test and train AI agents and models against diverse attack techniques. Strikes enables iterative improvement of AI resilience through repeated exposure to offensive tactics.
Spyglass: An AI red teaming toolkit for probing deployed AI applications. Spyglass allows continuous auditing and vulnerability assessment of live AI models, integrating adversarial machine learning research to detect issues such as prompt injection, model bypasses, and data poisoning.
Crucible: An AI hacking sandbox where practitioners can test, learn, and advance their AI red team skills in a safe, controlled environment. Crucible supports experimentation and skill development for both individuals and organizations.
Rigging Library: An open-source agent development library with features like execution tracing, API tool integration, HTTP endpoint modeling, and multi-agent orchestration, enabling flexible and complex AI agent behaviors.
Dyana: An open-source eBPF sandbox for dynamic testing of machine learning models and other files, providing observability into GPU memory usage, filesystem interactions, network requests, and security events. Dyana is designed to detect backdoors and supply chain risks in AI models before deployment.
Key Capabilities
Offensive AI simulation and red teaming
Automated vulnerability discovery and exploitation for AI systems
Dynamic testing and evaluation of AI and machine learning models
Real-world scenario generation for AI agent training
Supply chain security and model observability
Open-source tooling for agent development and model validation
Investors
Series A Funding: $14 million raised in early 2025
Lead Investor: Decibel Partners
Other Investors: Next Frontier Capital, In-Q-Tel (IQT), Sands Capital, Indie VC, Aviso Ventures, Jon Oringer, and others
Valuation: Not publicly disclosed
Company Type: Privately held
Notable Clients
Dreadnode is trusted by enterprises, government agencies, and frontier AI model providers. While specific client names are not publicly listed, the company’s solutions are used by organizations seeking to advance their offensive security capabilities and ensure the robustness of AI deployments.
